Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

OpenAI and Meta warn US judicial committee that deepfakes threaten litigation evidence

AI summary

Representatives of OpenAI and Meta told a US Judicial Conference Advisory Committee on Evidence Rules meeting on Oct. 15 that AI-driven deepfakes pose a serious threat to the reliability of evidence in litigation, IT之家 reported, citing Reuters. OpenAI policy adviser Shira Anderson said AI-generated images keep improving and may become impossible to identify scientifically, while Meta product management director Devin Norder said humans can no longer distinguish real media from output of recent frontier AI models and that producing such media now takes negligible time and cost. The report also said experts warn courts are finding it harder to resist AI-generated false evidence and that several AI companies add visible or invisible watermarks, though no universal detection method exists; still, in a survey of 931 US judges only 15 had faced a credibility challenge over suspected deepfakes.

In-depth insight

Two major AI companies publicly told a federal evidence-rules forum that deepfakes threaten litigation evidence, while the same report's survey of judges shows such disputes remain rare so far.

IT之家 10 月 20 日消息,据路透社报道,OpenAI 与 Meta 代表在当地时间本月 15 日的美国司法会议证据规则咨询委员会会议上均表示,人工智能驱动的深度伪造将严重威胁法律诉讼中的证据可靠性。

OpenAI 政策顾问 Shira Anderson 表示:

AI 生成的图像质量正不断提升,我认为未来将无法通过科学手段来识别它们。

Meta 产品管理总监 Devin Norder 则称:

人类已无法分别真实媒体与至少最近几代前沿 AI 模型生成的媒体。而目前制作此类媒体所需的成本和时间确实微乎其微。

图源:Pexels

专家警告称,法院抵御人工智能生成的虚假证据的难度正与日俱增。目前多家人工智能企业已向 AI 生成媒体添加可见或不可见的水印,但目前尚无检测这些水印的通用方法。

不过,深度伪造在司法界的实践中尚属少见:在一份 931 名美国法官参与的调查中,只有 15 名法官遭遇过因涉嫌深度伪造质疑证据可信度的情况。

#OpenAI#Meta#US Judicial Conference#Advisory Committee on Evidence Rules

Original · 1 min readRead the original
Group chat

No comments yet. Start the conversation.